feat: add Vara Network (VARA, coin type 913) and two-byte SS58 prefix support

[HAPPYS1NGH]

Closes #417

Changes

• SS58 helpers (utils/dot.ts) now encode/decode two-byte network prefixes (64–16383) — Vara uses prefix 137 — any Substrate chain with a high prefix can now be added in a few lines.
• Decoder checks the decoded length — 1.1.4 accepts a valid address with bytes appended after the checksum and returns the same key — malformed SS58 input is now rejected.
• createDotAddressEncoder/Decoder throw on an out-of-range prefix at creation — previously new Uint8Array([type]) silently wrapped it — misconfiguration fails loudly.
• Added vara (coin type 913, SS58 prefix 137), registered in coins.ts, coders.ts, coinTypeToNameMap.ts and the supported-coins doc — Vara addresses resolve through ENS.
• Tests: Vara encode/decode vectors, rejection of a Polkadot address, a bad checksum and a truncated address, plus a dot regression test for trailing bytes.

Notes

• References: SLIP-44 913 VARA Vara Network; ss58-registry prefix: 137, network: "vara"; SS58 address format.
• Test vectors: the Substrate dev accounts Alice and Bob plus a live Vara account, all checked against @polkadot/util-crypto encodeAddress(pubkey, 137).
• Verified: bun test 1368 pass / 0 fail, bun run build passes. dot and ksm vectors are unchanged.
• Cross-check (not committed): compared the built helpers to @polkadot/util-crypto across every one-byte prefix, the 63/64 and 16383 boundaries, 137 and 40 random two-byte prefixes, in both directions (11,391 checks, 0 failures). Every single-character corruption of a Vara address is rejected.
• No new dependencies. It reuses @noble/hashes and the existing base58 helpers.
• Not touched: the docs generator also re-pads the EVM table, which is already out of date on master; I left that out to keep the diff focused.

🤖 Generated with Claude Code